Imported from projectious-work/aibox (context/templates/processkit/v0.22.0/context/skills/documents/xlsx-modeling/SKILL.md). Install upstream with npx skills add projectious-work/aibox --skill xlsx-modeling. Copyright stays with the author.

XLSX Modeling

Intro

Use openpyxl to generate Excel files programmatically. Separate inputs, calculations, and outputs into distinct sheets or named ranges — this makes models auditable and reduces errors. Write cell values as Python data and let Excel formulas do the math; never pre-calculate values that Excel should own.

Overview

Setup

# /// script
# dependencies = ["openpyxl"]
# ///
from openpyxl import Workbook, load_workbook
from openpyxl.styles import Font, PatternFill, Alignment, Border, Side
from openpyxl.utils import get_column_letter, column_index_from_string

wb = Workbook()
ws = wb.active
ws.title = "Summary"

Install with uv add openpyxl or pip install openpyxl.

Writing data

# Single cell
ws["A1"] = "Revenue"
ws["B1"] = 1_250_000

# Row at a time
ws.append(["Q1", 312500, "=B2/B$1"])    # mix of values and formulas
ws.append(["Q2", 340000, "=B3/B$1"])

# From a list-of-lists
data = [["Name", "Score"], ["Alice", 92], ["Bob", 87]]
for row in data:
    ws.append(row)

Rows are 1-indexed. ws["A1"] and ws.cell(row=1, column=1) are equivalent. Use get_column_letter(n) to convert column numbers to letters.

Formulas

Write Excel formula strings directly. openpyxl writes the formula; Excel evaluates it on open.

ws["C2"] = "=SUM(B2:B5)"
ws["D2"] = "=AVERAGE(B2:B5)"
ws["E2"] = "=IF(B2>1000, \"High\", \"Low\")"
ws["F2"] = "=VLOOKUP(A2, Lookup!A:B, 2, FALSE)"

Set data_only=True when reading a file to get cached values instead of formula strings: load_workbook("file.xlsx", data_only=True).

Named ranges

from openpyxl.workbook.defined_name import DefinedName
from openpyxl.utils import quote_sheetname, absolute_coordinate

# Define a named range
ref = f"{quote_sheetname('Summary')}!{absolute_coordinate('B2:B13')}"
wb.defined_names["MonthlyRevenue"] = DefinedName("MonthlyRevenue", attr_text=ref)

# Reference it in a formula
ws["C2"] = "=SUM(MonthlyRevenue)"

Use named ranges for values referenced across multiple sheets — it makes formulas readable and survives row/column insertions.

Formatting

# Font
ws["A1"].font = Font(name="Calibri", bold=True, size=12, color="1F497D")

# Fill
ws["A1"].fill = PatternFill(fill_type="solid", fgColor="4472C4")

# Alignment
ws["A1"].alignment = Alignment(horizontal="center", vertical="center",
                                wrap_text=True)

# Border
thin = Side(style="thin")
ws["A1"].border = Border(left=thin, right=thin, top=thin, bottom=thin)

# Number format
ws["B2"].number_format = "#,##0.00"          # 1,234.56
ws["C2"].number_format = "0.0%"              # 12.3%
ws["D2"].number_format = '"$"#,##0'          # $1,234
ws["E2"].number_format = "YYYY-MM-DD"        # 2026-04-08

Apply styles to header rows and totals rows consistently. Use a defined NamedStyle object to apply a style set to multiple cells:

from openpyxl.styles import NamedStyle
header_style = NamedStyle(name="header")
header_style.font = Font(bold=True, color="FFFFFF")
header_style.fill = PatternFill(fill_type="solid", fgColor="2E74B5")
wb.add_named_style(header_style)
ws["A1"].style = "header"

Column widths and row heights

ws.column_dimensions["A"].width = 20
ws.column_dimensions["B"].width = 14
ws.row_dimensions[1].height = 30   # points

# Auto-width approximation (openpyxl has no auto-fit)
for col in ws.columns:
    max_length = max(len(str(cell.value or "")) for cell in col)
    ws.column_dimensions[col[0].column_letter].width = min(max_length + 4, 60)

Charts

from openpyxl.chart import BarChart, Reference

chart = BarChart()
chart.type = "col"
chart.title = "Quarterly Revenue"
chart.y_axis.title = "Revenue ($)"
chart.x_axis.title = "Quarter"

data_ref = Reference(ws, min_col=2, min_row=1, max_row=5)
categories = Reference(ws, min_col=1, min_row=2, max_row=5)
chart.add_data(data_ref, titles_from_data=True)
chart.set_categories(categories)
chart.shape = 4         # rectangle (no 3D)

ws.add_chart(chart, "D2")

Chart types: BarChart, LineChart, PieChart, AreaChart, ScatterChart, BubbleChart. Set chart.shape = 4 (flat) for all chart types — 3D charts are harder to read.

Freeze panes and filters

ws.freeze_panes = "B2"       # freeze row 1 and column A
ws.auto_filter.ref = "A1:D100"   # enable autofilter on header row

Multiple sheets

ws_data  = wb.create_sheet("Raw Data")
ws_calc  = wb.create_sheet("Calculations")
ws_out   = wb.create_sheet("Output")

# Reorder: move Output to first position
wb.move_sheet("Output", offset=-wb.index(wb["Output"]))

# Remove the default sheet if unused
del wb["Sheet"]

Saving

wb.save("report.xlsx")

from io import BytesIO
buf = BytesIO()
wb.save(buf)
buf.seek(0)

Gotchas

Agent-specific failure modes — provider-neutral pause-and-self-check items:

• Pre-calculating values that Excel formulas should own. Writing ws["C2"] = sum(data) instead of ws["C2"] = "=SUM(B2:B100)" makes the spreadsheet static — users cannot change inputs and see results update. Write formulas for all derived values; let Excel own the calculation.
• Not separating inputs, calculations, and outputs into distinct sheets. A single sheet that mixes raw data, intermediate calculations, and final results becomes impossible to audit. Organize into at minimum an "Inputs" sheet and an "Output" or "Summary" sheet; keep raw data on its own sheet untouched.
• Using hard-coded cell references in formulas instead of named ranges. A formula like =SUM(Sheet1!B2:B13) breaks silently if a row is inserted before row 2. Named ranges (e.g. MonthlyRevenue) survive structural changes and make formulas self-documenting.
• Omitting number formats on numeric cells. Raw numbers without formatting (1234567.89) are harder to read than formatted ones ($1,234,567.89). Set .number_format on all monetary, percentage, and date cells before saving.
• Reading a file with data_only=False expecting cached values. When data_only=False (the default), openpyxl returns formula strings, not computed values. To read the last computed result, open with load_workbook("file.xlsx", data_only=True). If the file has never been opened in Excel, the cached values will be None.
• Adding 3D chart types. 3D bar and pie charts are visually misleading and harder to read than flat equivalents. Set chart.shape = 4 (rectangular) and never use 3D variants in generated output.
• Not freezing header rows on data sheets. A data sheet without ws.freeze_panes = "A2" forces users to scroll to remember column labels. Always freeze at least the header row on any sheet with more than one screen of rows.

Full reference

openpyxl vs xlsxwriter

Use openpyxl when you need to read and modify an existing file. Use xlsxwriter when producing large files from scratch with rich chart formatting and you don't need to read back the result.

Large files (streaming write)

from openpyxl import Workbook

wb = Workbook(write_only=True)
ws = wb.create_sheet()
ws.append(["Name", "Score"])    # header
for name, score in dataset:
    ws.append([name, score])
wb.save("large.xlsx")

write_only=True uses constant memory — cells cannot be read back after writing.

Conditional formatting

from openpyxl.formatting.rule import ColorScaleRule, DataBarRule

rule = ColorScaleRule(
    start_type="min", start_color="F8696B",
    mid_type="percentile", mid_value=50, mid_color="FFEB84",
    end_type="max", end_color="63BE7B",
)
ws.conditional_formatting.add("B2:B100", rule)

Data validation (dropdown lists)

from openpyxl.worksheet.datavalidation import DataValidation

dv = DataValidation(type="list", formula1='"Yes,No,N/A"', allow_blank=True)
ws.add_data_validation(dv)
dv.add("C2:C100")

Password protection

ws.protection.sheet = True
ws.protection.password = "readonly"
ws.protection.enable()